The talk “Neuronflow: an event-driven processor architecture for low power AI at the edge” will be presented March the 6th, at 11:30, in room I-105, moderated by Pedro Diniz (DEI).
Abstract:
This presentation provides an in-depth overview of the GrAIcore Neural Processing Unit (NPU) architecture developed by Snap Inc. for Augmented Reality applications, emphasizing the role of sparsity in achieving significant power and performance improvements compared with traditional architectures. The discussion will focus on the necessity of redesigning traditional computer architectures to adopt an event-driven execution model, which is critical for harnessing the benefits of sparsity. Additionally, the talk will explore advanced training optimization techniques that enhance network activation sparsity, enabling the full potential of the event-driven multicore architecture to be realized. This holistic approach to architectural and algorithmic design is essential for leveraging the unique advantages of the GrAIcore NPU in real-world applications.
About the Speaker:
Orlando Moreira is Chief Computer Architect and Senior Manager at Snap Inc. in Eindhoven, Netherlands. His expertise encompasses computer architecture, edge AI, embedded systems, real-time systems, and data flow methodologies. Before his tenure at Snap Inc., Moreira held the position of Chief Architect at GrAI Matter Labs, where he was responsible for the compute architecture and software development kit (SDK) roadmaps.
Before, he worked for Philips Research, ST-Ericsson, Ericsson, and Intel (where he was group leader for programming and core tools – compiler, simulator, debugger and hw generation).
Throughout his career, Moreira has contributed peer-reviewed to the field of computer architecture, particularly in the areas of embedded and cyber-physical systems, as well as the field of real-time design, modeling, and analysis .
He holds a PhD in Electrical Engineering from the University of Eindhoven.